December 11, 2004

The Marriage Gap

My sequel to "Baby Gap" will be up on VDARE late Sunday night. It's hot stuff ... if you are interested in understanding what drives the blue-red gap in election results. I've now found a demographic factor that correlates even better with Bush's share of the vote by state than total lifetime fertility of white women does. (It's not a terribly different factor than fertility, so don't expect a huge surprise.)

When you put the new mystery factor together with fertility in a simple multiple regression model, you get an r-squared of 88%, which is bizarrely high. That means that if you have just these two demographic measures for each state (in fact, those just for the white residents, weirdly enough), you can come up with a model where only 12% of the variation in Bush's share is unaccounted for. And it worked almost as well in 2000.

Think about all the reasons that pundits gave for why Bush or Kerry would do well in a particular state -- the strength of the state's economy, whether or not the candidates platforms would be good or local interests, the popularity of Gov. Schwarzenegger in California or the unpopularity of the scandal plagued GOP in Illinois, or the number of visits the candidates paid to the state, or yada yada yada. All trivial, accounting in sum for 12% of the variation, compared to the two big demographic factors that nobody mentioned. Granted, they are still very important, but you can see why all the campaign resources were poured into the small number of battleground states where the demographic factors put them on the cusp.

Most strangely, the racial makeup of the state doesn't much matter in this model. Because blacks gave 88% of their vote to Kerry, while whites gave him only 41%, common sense says that the percentage of voters in a state who are black would play an important role in determining Bush's share of the vote. Yet, you can get to an r-squared of 88% without inputting the black share of the state's voters. Remarkably, how large a share of the state's populace is black apparently influences the state's white fertility and the white mystery demographic factor enough to account for the varying black influences on the state's voting outcome. (I will, however, eventually input each state's racial makeup and see if that makes the model even more accurate. But the fewer factors in a model the better, on the whole. You want to make it as simple as possible, but no simpler, as Einstein might have of said.)

My published articles are archived at -- Steve Sailer

No comments: